Interesting paper. Its now "old" in that it was published in 1999.
Couple of quick comments:
I have taught Software Management using Brook's Mythical Many Month as a text. I believe in it. It may be wrong, but its held up for 25 years, and still rings true to me.
Nothing about the Internet for communications, as claimed by Cathedral and Bazaar (CatB), really changes the essence of Brook's observation. Nine women can not deliver a baby in one month. Some things just don't scale. What we know from thirty or forty years of software engineering is that technology is not the problem that makes large projects "hard", its the human to human communications.
I believe that fewer good people make projects run faster, not more. And I believe that if you are going to have many people, you need large numbers of eyeball to eyeball time. With perhaps two or three percent of that time over beer.
On "Given enough eyeballs, all bugs are shallow"
I don't buy this one at all. For many, perhaps even most, bugs, more eyeballs can help. But complex interactions, race conditions, inverted interrupts and other stuff that happens in real world systems as you push for performance and scalability are hard. They are essentially hard. It takes a lot of time for an experienced artisan to get into the zone to start to understand them. The world simply does not have 'enough eyeballs' attached to expert artisans, so some bugs are hard.